search

dhm2013724 / yolov2_xilinx_fpga

A demo for accelerating YOLOv2 in xilinx's fpga pynq/zedboard
MIT License
739 stars 229 forks source link

what is weight and beta offset ? #18

Open Thilanka97 opened 5 years ago

Thilanka97 commented 5 years ago

@dhm2013724 Hey, I am trying to understand how your implementation works and I do not understand what weight offset and beta_offset is. Also What it does? Please explain. Also is inputQ, outputQ, weightQ and betaQ mean quantized input, quantized output, quantized weight and quantized beta? do you input both quantized weight, beta files and 32fp weight, beta files to the design? (I am confused because there are number of weight and beta files) what is the difference between weight and weightQ? how are they used in the design ?

Thanks in advance!

liutongxue123 commented 5 years ago

@dhm2013724 Hey, I am trying to understand how your implementation works and I do not understand what weight offset and beta_offset is. Also What it does? Please explain. Also is inputQ, outputQ, weightQ and betaQ mean quantized input, quantized output, quantized weight and quantized beta? do you input both quantized weight, beta files and 32fp weight, beta files to the design? (I am confused because there are number of weight and beta files) what is the difference between weight and weightQ? how are they used in the design ?

Also what is Tr and Tc? and mloops, nloops,rloops and cloops?

Thanks in advance!

我也在修改此代码 方便一起交流下么

Thilanka97 commented 5 years ago

@liutongxue123 nice! do you understand what is the difference between weight and weightQ? how are they used in the design ? what inputQ, outputQ, weightQ and betaQ are?

Thilanka97 commented 5 years ago

@dhm2013724 how did you select the Tn, Tm, Tr and Tc <2,32,26,26> values?

Thanks in advance!

Thilanka97 commented 5 years ago

@JinzhongHe I read some. But I can not get the full pdf versions of these. I can see only the abstract. [6] A Dynamic Multi-precision Fixed-Point Data Quantization Strategy for Convolutional Neural Network [7] Caffeine: Towards Uniformed Representation and Acceleration for Deep Convolutional Neural Networks

can u send m if you have full versions. Also from which paper can you get an idea about inputQ, outputQ, weightQ and betaQ?

Thanks in advance!

Thilanka97 commented 5 years ago

@JinzhongHe Sorry I dont use QQ

Thilanka97 commented 5 years ago

@JinzhongHe hey I can not download the weight files from the baidu cloud. Can u send it to me if you have them downloaded. email- thiltt97@gmail.com

Thanks in advance!